Deep learning for 40 MHz scouting with Level-1 trigger muons for CMS at LHC run-3

The muon track finder of the CMS experiment at the Large Hadron Collider uses custom FPGA-based processors to identify muons and measure their momentum for a fast Level-1 trigger selection. A 40 MHz scouting system at CMS will provide fast statistics for detector diagnostics, alternative luminosity measurements, and new analysis possibilities.

Deep learning is a subfield of machine learning algorithms that uses multiple hidden neural layers to extract relevant features from raw inputs. Previous studies have demonstrated the potential of deep learning in many areas of particle physics. The purpose of this study is to analyse the performance of deep learning algorithms to recalibrate the muon track parameters (transverse momentum, η and 𝜑) for the best resolution. Deep learning regression models are compared against simple linear fits. The performance of these models is evaluated and compared.